LANGUAGE AND LINGUISTICS 9.1:101-126, 2008
2008-0-009-001-000252-1
The Encoding of Referential Properties in the
Chinese Nominal*
Joanna Ut-seong Sio
Hong Kong Polytechnic University
By manipulating the positions of modifiers, it can be shown that in Chinese,
referential properties like specificity and definiteness are related to some structure
higher than the numeral. This structure is identified as the Specificity Phrase (SP),
the projection of which gives rise to specificity. The feature specification of the S
head determines the definiteness interpretation of a noun phrase. Not all Chinese
noun phrases can be interpreted as definite. The restrictions in definiteness interpretation are modeled by a theory in which definiteness is regulated by an Agree
(Chomsky 2000, Pesetsky & Torrego 2004) relation between the Specificity Phrase
and the Classifier Phrase. Every referential interpretation is unambiguously linked
to a corresponding nominal structure with a certain feature specification.
Key words: definiteness, specificity, classifiers, modifiers
1. Introduction
In Chinese, there are no articles indicating the definiteness value of a noun phrase.
As a consequence, the referential interpretations of some Chinese noun phrases are
relatively flexible in the sense that one surface form can have many possible referential
interpretations. For instance, bare nouns in Mandarin and [Cl-N] phrases in Cantonese
can be interpreted as either definite or indefinite (Cheng & Sybesma 1999); indefinite
noun phrases are always ambiguous between a specific and a non-specific reading. To
complicate the picture, it has been observed that only definite noun phrases can appear
in the subject position in Chinese (see Chao 1968, Li & Thompson 1981, Lee 1986,
among others). The possibility arises that the referential interpretations of these flexible
noun phrases are not determined by factors internal to the noun phrase. The interpretation
is either determined by the discourse context, or is dependent on the position of the
noun phrase in the sentence. However, a closer look at the forms and interpretations
*
The author would like to thank Boban Arsenijevic, Lisa Cheng, Rint Sybesma and Luis Vicente
for comments and criticisms along the development of the paper. All errors are solely the
responsibility of the author.
Joanna Ut-seong Sio
co-relates of these flexible Chinese noun phrases, together with dialectal comparison,
shows that the referential interpretations of these noun phrases are not as flexible as
they seem to be at first sight. For instance, [Cl-N] phrases in Mandarin and bare nouns
in Cantonese can never be interpreted as definite, regardless of contexts or sentence
positions. This shows that neither a context-based nor a sentence-position-based account
is adequate in accounting for the variations. What then determines the referential
properties of Chinese noun phrases?
There is evidence to show that referential properties of Chinese noun phrases have
a close-tie with nominal structure. It has been observed that in Chinese, the positions of
modifiers inside a noun phrase interact with referential properties (Huang 1982, Cheung
1989, Zhang 2006). Briefly, when a modifier appears to the immediate left of the classifier
or the numeral, the noun phrase is, if grammatical, obligatorily referential (specific and
in some cases definite). Placing a modifier between the classifier and the noun, on the
other hand, has no effects on the referential interpretation of the noun phrase. This
suggests that the Chinese noun phrase has more structure than meets the eye. Some
structure related to referentiality is present in the noun phrase, located higher than the
numeral. I identify such referentially related structure as the Specificity Phrase (SP).
The S head always contains the feature SPECIFIC, giving rise to a specific interpretation.
Since both definite and indefinite noun phrases can be specific, definiteness is treated as
a value of the SPECIFIC feature.
Not all Chinese noun phrases can be interpreted as definite. What this implies is
that the assignment of the definiteness value to the SPECIFIC feature is not free, but is
dependent on some other factors. In this paper, I adopt the idea that referential properties
in the nominal are regulated by the interplay between two referentially related layers in
the nominal (Szabolcsi 1994, Campbell 1996, Hoekstra & Hyams 1996, Brugè 2002,
among others). Further assuming that definiteness is encoded in the Classifier Phrase
(Cheng & Sybesma 1999), I take it that the Specificity Phrase and the Classifier Phrase
are the two main players in regulating the referential properties of Chinese nominals.
The two layers interact via Agree (Chomsky 2000, Pesetsky & Torrego 2004).
This paper is organized as follows. In §2, I present definitions of terms that I would
be using throughout the paper. In §3, I argue that referential properties of Chinese noun
phrases are neither determined by the context nor by the position of the noun phrase
within a sentence. In §4, I present evidence to show that referential properties of
Chinese noun phrases are sensitive to nominal structure. In particular, there is some
structure to the left of the numeral that is related to referential properties, which I
identify as the Specificity Phrase (SP). In §5, I propose a theory in which referential
properties are regulated by an Agree relation (Chomsky 2000, Pesetsky & Torrego 2004)
between the Specificity Phrase and the Classifier Phrase. In §6, I provide structures for
102
The Encoding of Referential Properties in the Chinese Nominal
[Dem-(Nume)-Cl-N] phrases, [Nume-Cl-N] phrases, [Cl-N] phrases, as well as bare
nouns in the two-layer system I propose, explaining how different interpretations can
arise. In §7, I discuss the licensing of different types of modifiers. Section 8 concludes
the paper.
2. Definitions and assumptions
In this section, I provide definitions for notions that I shall be using throughout the
paper.
2.1 Types of modifiers
Chinese modifiers come in two types. They can either be bare, as in (1) or they can
be followed by a modification marker, as in (2). The following examples are in
Cantonese.1
(1) [hung4-sik1]
red-color
(2) [hung4-sik1
red-color
syu1
book
ge3]
MARKER
syu1
book
For ease of exposition, I refer to the type of modifier in (1) as bare modifiers and
the type in (2) as marker modifiers.
2.2 Definiteness and specificity
I adopt the view that definiteness and specificity can be defined in terms of
membership of discourse (Arsenijevic 2008). Briefly, there are two relevant discourses
in every exchange viewing from the perspective of the speaker, the speaker’s discourse
and the shared discourse. A definite noun phrase is used to refer to an object the referent
of which is in the shared discourse. A specific noun phrase is used to refer to an object
the referent of which is in the speaker’s discourse, but not necessarily in the shared
discourse. Definiteness and specificity are thus defined as follows:
(3) A noun phrase is definite if its referent is present in the shared discourse.
(4) A noun phrase is specific if its referent is present in the speaker’s discourse.
1
Jyutping is used as the romanization scheme for the Cantonese data in this paper.
103
Joanna Ut-seong Sio
Since the shared discourse includes the speaker’s discourse, all definite noun phrases
are necessarily specific.
3. Not everything can be definite
If it were the case that Chinese does not encode referentiality in the nominal
structure, one of the possibilities is that the referential value of a Chinese noun phrase is
determined by contextual information. In other words, any interpretation, as long as it is
compatible with the context, should be available. In this section, we will focus on the
interpretation of definiteness. Consider the following Cantonese example:
jiu3
maai5
(5) ngo5 jat1deng6
I
surely
must buy
‘I must buy a bag/the bag.’
go3
CL
doi2
bag
ga3
le3
SFP
SFP
go3doi2 in (5) is ambiguous. It can be understood in a way that what the speaker
demands is to buy a particular bag that both the speaker and the hearer know about, thus
a definite reading for go3doi2. (5) can also be understood in a way that the speaker is
just demanding to buy a bag, any bag would do, thus, an indefinite reading for go3doi2.
In other words, go3doi2 can be interpreted as either definite or indefinite, depending on
the context.
However, there is evidence to suggest that a purely context-based account cannot
be right. Even though referential interpretation of noun phrases is freer in Chinese, not
all noun phrases can be interpreted as definite. Cheng & Sybesma (1999) observe that
the definiteness interpretation of noun phrases varies across Chinese dialects, as illustrated
in the following table:
(6)
Cantonese
Mandarin
[Nume-Cl-N]
Indefinite
Indefinite
[Cl-N]
Definite/Indefinite
Indefinite
Bare noun
Indefinite
Definite/Indefinite
A Mandarin [Cl-N] phrase can never be interpreted as definite no matter how the
context manipulates. The same applies to Cantonese bare nouns. [Nume-Cl-N] phrases in
both dialects are always indefinite. No context can change that. If referential properties are
contextually determined, such restrictions in interpretation are unexpected. The variation
across dialects argues against the view that the interpretation of definiteness in Chinese
is merely context-dependent. The context is only relevant when a certain referential
104
The Encoding of Referential Properties in the Chinese Nominal
interpretation is within the ‘capacity’ of the noun phrase in question.
It has been widely observed that only definite noun phrases can appear in the
subject position in Chinese (see Chao 1968, Li & Thompson 1981, Lee 1986, among
others). For instance, even though a [Cl-N] phrase in Cantonese can be interpreted as
either definite or indefinite (as illustrated in (5)), a [Cl-N] phrase in the subject position
can only be interpreted as definite. This is illustrated in (7).
(7) zek3
gao2 jiu3
gwo3 mai5lou6
CL
dog want cross road
‘The dog wants to cross the road.’ NOT: ‘A dog wants to cross the road.’
This seems to suggest that sentential position can affect the definiteness of a noun
phrase. Nevertheless, this is only apparent. The subject position cannot force a definite
reading on just any noun phrase. For a noun phrase that cannot be interpreted as definite
to begin with, putting it in the subject position would only lead to ungrammaticality. For
instance, bare noun phrases can be interpreted as definite in Mandarin but not in
Cantonese. Putting a Cantonese bare noun phrase in the subject position would not
induce a definite reading on the bare noun phrase, but, rather, lead to ungrammaticality.
This is illustrated in (8).
(8) a.
gǒu yào
guò
mǎlù
(Mandarin)
dog want cross road
‘The dog wants to cross the road.’
gwo3 mai5lou6 (Cantonese)
b. * gao2 jiu3
dog want cross road
Intended reading: ‘The dog wants to cross the road.’
The contrast between (8a) and (8b) shows that the referential interpretation of
Chinese noun phrases cannot be solely determined by their sentential positions.
4. Structural sensitivity
In this section, we provide evidence to show that the encoding of specificity and
definiteness in Chinese correlates with structural properties within the noun phrase. In
particular, we show that the variation of the position of modifiers in a noun phrase
co-relates with specificity and definiteness.
105
Joanna Ut-seong Sio
4.1 Specificity
Huang (1982) observes the following contrast in existential sentences based on
Mandarin data (example (9b) and (9c) are from Huang (1982:64); I added example (9a)
for comparison):2
(9) a.
yǒu
[sān
běn shū]
zài
zhèr
have
three CL
book at
here
‘There are three books here.’
b. yǒu
[sān
běn zhāngsān de
shū]
zài
Zhangsan MARKER book
at
have
three CL
‘There are three books of Zhangsan’s here.’
c. * yǒu
[zhāngsān de
sān
běn shū]
zài
book
at
have
Zhangsan MARKER three CL
Intended reading: ‘There are three books of Zhangsan’s here.’
zhèr
here
zhèr
here
In (9a), there is no modifier and the [Nume-Cl-N] phrase can appear in existential
sentences. When a modifier is placed between the classifier and the noun,
[Nume-Cl-marker modifier-N], the resulting noun phrase can also appear in existential
sentences, as in (9b). However, when the modifier is placed to the left of the numeral,
[marker modifier-Nume-Cl-N], the resulting noun phrase is banned from existential
sentences. Huang (1982) suggests that the ungrammaticality of (9c) is related to the
referential/specific nature of the noun phrase in (9c).
Zhang (2006) points out that the type of noun phrase in (9b), [Nume-Cl-marker
modifier-N], is ambiguous between a specific and a non-specific reading, while the type
of noun phrase in (9c), [marker modifier-Nume-Cl-N] is obligatorily specific. Although
Zhang (2006) does not consider unmodified [Nume-Cl-N] phrases like (9a), it should be
noted that [Nume-Cl-N] phrases are also ambiguous between a specific and a non-specific
reading.
These observations can be schematized as follows:
(10) a. [marker modifier-Nume-Cl-N] Æ specific
b. [Nume-Cl-(marker modifier)-N] Æ specific/non-specific
In what follows, I shall provide one more piece of evidence to support the
observations in (10). The specific and non-specific distinction of indefinites is easily
detected in contexts that involve verbs of propositional attitude (e.g. want, believe, hope,
2
Pīnyīn is used as the romanization scheme for the Mandarin data.
106
The Encoding of Referential Properties in the Chinese Nominal
intend, etc.) (Lyons 1999). The specific and non-specific contrast can be made more
prominent by using an adverb like ‘randomly’. The interpretation that a specific noun
phrase carries is contradictory to the meaning of ‘randomly’, as manifested in the
senseless sentence ‘#I am going to randomly find three specific students.’ In order
words, only non-specific noun phrases should be compatible with ‘randomly’. The
compatibility with ‘randomly’ can then be used as a test for specificity of noun phrases.
In Mandarin, the word for ‘randomly’ is suíbiàn. In Cantonese, it is si6daan6. In what
follows, I shall use suíbiàn and si6daan6 to show that [Nume-Cl-(marker modifier)-N]
phrases can be either specific or non-specific while [marker modifier-Nume-Cl-N] phrases
are obligatorily specific.
Mandarin: [Nume-Cl-(marker-modifier)-N]
(11) wǒ xiǎng suíbiàn
zhǎo
I
want randomly look-for
[sān
ge
(chōuyān de)
xuéshēng]
smoke
MARKER
student
three CL
‘I want to randomly find three students (who smoke).’
Mandarin: [marker modifier-Nume-Cl-N]
(12) # wǒ xiǎng suíbiàn
zhǎo
I
want randomly look-for
[chōuyān de
sān
ge xuéshēng]
MARKER three
CL
student
smoke
Intended reading: ‘I want to randomly find three students who smoke.’
Cantonese: [Nume-Cl-(marker modifier)-N]
wan2
(13) ngo5 soeng2 si6daan6
I
want
randomly look-for
1
3
go
(sik6jin1 ge3)
hok6saang1]
[saam
CL
smoke
MARKER
student
three
‘I want to randomly find three students (who smoke).’
Cantonese: [marker modifier-Nume-Cl-N]
wan2
(14) # ngo5 soeng2 si6daan6
I
want
randomly look-for
6
1
3
ge
saam1 go3
hok6saang1]
[sik jin
MARKER
three
CL
student
smoke
Intended reading: ‘I want to randomly find three students who smoke.’
[Nume-Cl-(marker modifier)-N] phrases are compatible with ‘randomly’ as in (11)
107
Joanna Ut-seong Sio
and (13) while [marker modifier-Nume-Cl-N] phrases are not, as in (12) and (14). This
shows that the latter is obligatorily specific.
To recapitulate, a [Nume-Cl-N] phrase can be specific or non-specific. Having a
modifier between the classifier and the noun does not affect the referential properties of
a [Nume-Cl-N] phrase. However, when a modifier appears to the left of the numeral, the
phrase is obligatorily specific. The schema in (10) is repeated here as (15):
(15) a. [marker modifier-Nume-Cl-N] Æ specific
b. [Nume-Cl-(marker modifier)-N] Æ specific/non-specific
4.2 Definiteness
Cheung (1989) observes that (i) [Cl-N] phrases can be definite in Cantonese but
not in Mandarin; (ii) It is possible to put bare modifiers like possessors, relative clauses
and color terms immediately preceding [Cl-N] phrases in Cantonese but not in Mandarin.
This is illustrated in (16), (17) and (18).
(16) Possessors:
syu1
a. ngo5 bun2
CL
book
I
‘my book’
b. * wǒ
běn
shū
CL
book
I
Intended reading: ‘my book’
(17) Color-terms:
syu1
a. hung4sik1 bun2
CL
book
red
‘the red book’
b. * hóng sè
běn
shū
CL
book
red
Intended reading: ‘the red book’
(18) Relative clauses:
go3
naam4zai2
a. daai3 ngaan5 geng2
CL
boy
wear glasses
‘the boy who wears glasses’
b. * dài
yǎnjìng
ge
nánháizi
CL
boy
wear glasses
Intended reading: ‘the boy who wears glasses’
108
(Cantonese)
(Mandarin)
(Cantonese)
(Mandarin)
(Cantonese)
(Mandarin)
The Encoding of Referential Properties in the Chinese Nominal
When there is a modifier preceding the [Cl-N] phrase in Cantonese, as in (16a),
(17a) and (18a), the resulting phrase can only be definite. Cheung (1989) makes a
connection between the two observations, suggesting that a Cantonese [Cl-N] phrase
allows a bare modifier appearing to its left because it is definite. In Mandarin, a [Cl-N]
phrase is only indefinite, thus, no bare modifiers can appear to its immediate left.
I think Cheung’s (1989) observation is essentially correct. In fact, his observation
can be extended to other modifier types as well as to other dialects. Bare modifiers such
as locative phrases and temporal nominals can also immediately precede definite [Cl-N]
phrases in Cantonese, as illustrated in (19a) and (19b) respectively:
(19) Locative phrases:
daat3
a. toi2-soeng6-min6
CL
desk-top-surface
‘the stain on top of the desk’
zik1
stain
Temporal nominals:
dip6
sung3
b. kam4 jat6
yesterday
CL
dish
‘the dish from yesterday’
In addition to Cantonese and Mandarin, there is also evidence from other Chinese
dialects to support the observation that only definite [Cl-N] phrases would allow a bare
modifier to immediately precede it. In Wenzhou, a [Cl-N] phrase can be definite (20a)
and a bare modifier can appear in its immediate left (20b). In Hailu Hakka and
Taiwanese Southern Min, a [Cl-N] phrase cannot be definite ((21a) and (22a)) and a
bare modifier cannot appear to its immediate left ((21b) and (22b)).
(20) Wenzhou:3
a. pa313/35
s33
CL
book
‘the book/a book’
b. 24 pa313/35 s33
CL
book
I
‘My book’
3
I am grateful to You Rujie for providing the Wenzhou data.
109
Joanna Ut-seong Sio
(21) Hailu Hakka:4
a. * pun33 shu51
CL
book
Intended reading: ‘the book’
b. * ngai55 pun33 shu51
CL
book
I
Intended reading: ‘my book’
(22) Taiwanese Southern Min:
a. * pun55 cheh33
CL
book
Intended reading: ‘the book’
cheh33
b. * gua55 pun55
I
CL
book
Intended reading: ‘my book’
As a first approximation, we can make the following generalization:
(23) A definite [Cl-N] phrase can license a bare modifier to its immediate left.
Note that definiteness alone is not enough to license bare modifiers. Mandarin bare
nouns can be definite, but bare modifiers of the types mentioned above (e.g. possessors,
relative clauses, locatives, etc.) cannot precede a definite bare noun in Mandarin, as
illustrated in (24):
(24) * wǒ
I
shū
book
(Mandarin)
This suggests that the presence of an overt classifier is crucial. Note the use of the
word overt here. It has been suggested that the Classifier Phrase has the individualizing
function (Borer 2005) and is related to, among others, the expression of number and
definiteness (Cheng & Sybesma 1999). I assume that when a bare noun is used
non-generically, the Classifier Phrase is present nonetheless.
If both definiteness and an overt classifier are required in order to license a bare
modifier to the left of a referential noun phrase, it means that [Dem-Cl-N] phrases
should also be able to license bare modifiers to the immediate left of the [Dem-Cl-N]
4
I am grateful for Feng-fan Hsieh for providing me the Hailu Hakka and Taiwanese Southern
Min data.
110
The Encoding of Referential Properties in the Chinese Nominal
phrase. This is borne out, as illustrated in (25):
(25) hung4sik1 go2
red
that
‘that red book’
bun2
CL
syu1
book
(Cantonese)
The generalization in (23) can be revised as follows:
(26) A definite noun phrase containing an overt classifier can license a bare
modifier to its left.
4.3 The specificity phrase
To take stock, we have observed that: (i) when a marker modifier appears to the
left of a [Nume-Cl-N] phrase, the phrase is obligatorily specific; (ii) when a bare
modifier appears to the left of a [Cl-N] phrase, if grammatical, the phrase is obligatorily
definite. I would first like to dismiss the possibility that it is the addition of the modifier
that ‘makes’ the [Nume-Cl-N] phrase specific or the [Cl-N] phrase definite. First of all,
[Nume-Cl-N] phrases can be specific without the modifier. Similarly, [Cl-N] phrases
can be definite without the modifier. The respective noun phrases do not need the
modifier to achieve specificity or definiteness. Secondly, as shown in the [Cl-N] cases,
only Chinese dialects that permit definite [Cl-N] phrases would allow a bare modifier to
immediately precede the [Cl-N] phrase. If it is the case that the bare modifier makes the
[Cl-N] phrase definite, such requirement is unexplained. It is, furthermore, unclear how
the modifiers can have an effect on the referential interpretation of noun phrases. Thus, I
conclude that the specificity of [marker modifier-Nume-Cl-N] phrases and the definiteness
of [bare modifier-Cl-N] phrases are not from the modifiers but from the modifiees. The
co-relation can be articulated as follows:
(27) a.
Only a specific [Nume-Cl-N] phrase would allow a marker modifier
appearing to its immediate left.
b. Only a definite [Cl-N] phrase would allow a bare modifier appearing to
its immediate left.
The observations in (27) can be explained if some structure relating to referential
properties is located to the left of the classifier/numeral and such structure hosts the
modifier. The presence of modifiers to the left of the classifier/numeral then indicates
the presence of such structure. I identify such structure as the Specificity Phrase (SP).
The function of SP is to encode referential properties.
111
Joanna Ut-seong Sio
At first glance, the Specificity Phrase can function in the following manner. The S
head always contains a SPECIFIC feature that gives rise to a specific interpretation. Thus,
it is present in all specific (definite included) noun phrases. If the SPECIFIC feature has a
[+def] value, the noun phrase is definite. If it has no value, the noun phrase is indefinite.
Non-specific noun phrases do not project the SP layer.
(28) specific definite / specific indefinite
SP
3
S´
3
S
NumeP/ClP
SPECIFIC [+def] / SPECIFIC [ ]
However, the picture has to be more complex. The SP layer alone is not enough to
account for all the variations of Chinese noun phrases. In Chinese, [Dem-(Nume)-Cl-N]
phrases are always definite; [Nume-Cl-N] phrases are always indefinite; [Cl-N] phrases
and bare noun phrases can be either definite or indefinite depending on the dialect. What
this means is that the assignment of definiteness values on the SPECIFIC feature ([+def]
or ∅) is not totally free. It is dependent on other factors. The proposal that I would like
to advance here makes use of the interaction between the Specificity Phrase and the
Classifier Phrase in determining the referential properties of the noun phrase. It has
been suggested that the classifier takes up many of the functions of articles (Cheng &
Sybesma 1999, Borer 2005). I take it as an indication that the classifier head is also
related to referential properties. In particular, the classifier head is where definiteness is
encoded (Cheng & Sybesma 1999). In other words, there are two layers in the Chinese
nominal that are related to referential properties, the Specificity Phrase and the Classifier
Phrase.
The idea that there are two referentially-related (R-related) layers in the noun
phrase is not novel. In fact, different proposals have been put forth bearing variations of
the same theme. These proposals arrive at the same conclusion via quite different routes.
I shall briefly mention some of the work below.
Szabolcsi (1994) proposes that there are two R-related layers in the nominal based
on Hungarian data: a DP that is the topmost FP and a DetP that is closer to the lexical
core. The D head hosts articles: a(z) ‘the’ or ∅ ‘a, some’. The articles act as
subordinators in the sense that they enable the noun phrase to function as arguments.
DetP determines both the quantification and definiteness of the noun phrase. It hosts
quantificational elements like minden ‘every’, kevés ‘few’ and semelyik ‘neither’ or the
112
The Encoding of Referential Properties in the Chinese Nominal
demonstratives e, eme, ezen ‘this’, ama, azon ‘that’. The DP layer and the DetP layer
are related in the following manner: Articles like a(z) ‘the’ or ∅ ‘a, some’ are selected
for D in agreement of the (in-)definiteness determined by Det.
Hoekstra & Hyams (1996), drawing a parallel between the clausal and nominal
domain, propose that there is a counterpart of tense in the nominal, call it X. The deictic
operator, representing the speech environment, binds a variable in X. This ‘interaction’
between the operator and X determines the referential properties of the DP as a whole.
(29) OP
D
NUM
[DP X
[NP N]]
Campbell (1996) proposes that in specific common noun phrases, there is a
specificity operator in SpecDP. The specificity operator binds the subject position of a
small clause, as shown in (30):
(30) [DP OPi the [SC [e]i thief]]
The specificity operator is a kind of DP-internal topic, which links the internal subject position (and the DP itself) to a referent identified previously in the discourse. The
specificity operator is generated in the subject of a lower functional projection, Article
Phrase (ArtP). The specificity operator, when overt, is realized by the demonstrative.
The operator (together with the demonstrative when overt) moves to a higher functional
projection, DP. The operator is covert when there is other material in the higher D head
such as the definite article.
Brugè (2002) proposes that the demonstrative in Spanish is generated in the
specifier position of a functional projection intermediate between the DP and the NP
and lower than all the functional projections containing APs.
(31) DP — APs — Demonstrative — N — complements
As illustrated in the Spanish examples below, the demonstrative can appear in
either the high SpecDP position as in (32a) or in the low SpecFP position in surface
structure, as in (32b) (the head noun is moved to some head position of a functional
projection between the DP and the APs, call it #P):
[DP estei
this
b. [DP el
the
(32) a.
[#P libroj
book
[#P libroj
book
[FP ti [NP tj ]]]]
[AP viejo [FP este [NP Nj]]]]]
old
this
113
Joanna Ut-seong Sio
In (32a), the demonstrative moves up to SpecDP to check its [+ref] and [+deictic]
feature with D, while in (32b), the adjective blocks the movement (assuming that
adjectives are maximal projections in the Spec of some functional projection, à la
Cinque (1994). The definite article is inserted to show that this position contains some
particular feature (i.e. [+ref]) to prevent it from being interpreted as existential. The
demonstrative then moves up at LF to check its feature.
Details aside, the various proposals above share the theme that there are two
R-related layers in a noun phrase and there is interplay between the higher R-related
layer and the lower R-related layer. The question is then what mechanism regulates the
two R-related layers in Chinese, the Specificity Phrase and the Classifier Phrase.
5. The proposal: Agree, definiteness and indefiniteness
The S head is the locus of specificity. The S head always contains the feature
SPECIFIC, which gives rise to a specific reading. Specific noun phrases can be either
definite or indefinite. A SPECIFIC feature that has a [+def] value gives rise to a definite
reading. A SPECIFIC feature that has no value, ∅, gives rise to a specific indefinite
reading by some default rule at LF. It is similar to treating tense as a TENSE feature, and
[+past] as a value of that feature. By assumption, the classifier is where definiteness is
encoded (Cheng & Sybesma 1999). The S head needs to seek the classifier to provide
the SPECIFIC feature with a definiteness value. I assume that there are two types of
classifiers. Classifiers can come out of the lexicon either carrying a [+def] value or
containing no value specification, ∅. Definite noun phrases are always specific. Indefinite
noun phrases can be either specific or non-specific. In other words, classifiers with a
[+def] value would require the projection of the SP layer in order to get interpreted.
Classifiers that contain no value specification can be interpreted with or without the SP
layer. The SP layer and the definite classifier ‘need’ each other, but for different reasons.
The S head with the SPECIFIC feature seeks a value. The definite classifier seeks to be
interpreted in the SP domain.
The two-way relationship between the S head and the definite classifier can be
implemented by an Agree relation (Chomsky 2000), in which the probe and goal ‘need’
different things from each other. The probe needs a value and the goal needs to be
interpreted in a proper domain. Pesetsky & Torrego’s (2004) formulation of Agree
provides such an option.
Pesetsky & Torrego’s (2004) argue for a modification of the formulation of the
agreement mechanism proposed in Chomsky (2000). They first of all differ from
Chomsky’s formulation in that they argue in favor of a view of Agree as feature sharing.
They define it as follows:
114
The Encoding of Referential Properties in the Chinese Nominal
(33) Agree
(i) An unvalued feature F (a probe) on a head H at a syntactic location α
(Fα) scans its c-command domain for another instance of F (a goal) at a
location β (Fβ) with which to agree.
(ii) Replace Fα with Fβ so the same feature is present in both locations.
They abandon the Valuation/Interpretability Biconditional:
(34) Valuation/Interpretability Biconditional (Chomsky 2001:5)
A feature F is uninterpretable iff F is unvalued.
If interpretability is concerned with whether a feature can be interpreted in the
semantic component and valuation is concerned with whether that particular feature has
been specified or not, interpretability should not be identical to valuation. One of the
consequences of having valuation and interpretability detached from each other is that it
gives rise to two more types of feature combinations:
Type of features (boldface = new type of combinations, taken from Pesetsky &
Torrego 2004):
(35) a. uF val uninterpretable, valued b. iF val interpretable valued
c. uF [ ] uninterpretable, unvalued
d. iF [ ] interpretable, unvalued
([ ] means unvalued, val means valued. u means uninterpretable while i
means interpretable.)
In Pesetsky & Torrego’s (2004) system, both uninterpretable and unvalued features
can act as probes. They give a plausible example of an interpretable unvalued feature
acting as a probe, namely, the T feature of the category Tense. They argue that, assuming
there is a distinct Tense node acting as the locus of semantic tense (following Chomsky
1957, Emonds 1976, 1978 and Pollock 1989), and in view of the fact that in many
languages, tense morphology is on the verb, this combination of affairs suggests that T
on the finite verb in such languages would bear an uninterpretable but valued T feature
that enters an Agree relation with an interpretable but unvalued T feature on Tense. (33)
schematizes the Agree relation.
(36) The relationship between Tense and the finite verb:
Agree
…Tense …
[v walked]…
…Tense … [v walked]…
uT [+past ]
iT [ ]
uT [+past]
iT [+past]
115
Joanna Ut-seong Sio
In the above case, T is the feature and [+past] is the value. The underlined value
represents the one that receives the value from the host feature via Agree. Pesetsky &
Torrego (2004) use the term instance to refer to the feature-location pair and the term
occurrence to refer to the distinct features that might undergo Agree. Thus, in (36), in
the pre-Agree state (the left side of the arrow), there are two occurrences of the T
features; in the post-Agree state (the right side of the arrow), there are two instances.
Deletion applies then to the uninterpretable instance of the feature T.
Coming back to Chinese, the S head, being the locus of specificity, has an iSPECIFIC
feature that is unvalued, [ ]. The iSPECIFIC feature always gives rise to a specific
interpretation, but is unvalued for definiteness. The iSPECIFIC [ ] feature on the S head,
the probe, has to agree with a goal in his c-command domain. Two types of classifiers
can be drawn from the lexicon. It either has an uSPECIFIC [+def] value, or it can come
with no SPECIFIC feature/definiteness value at all.5
(37) S: iSPECIFIC [ ] = interpretable SPECIFIC feature but unvalued
Cl: uSPECIFIC [+def] or ∅ = uninterpretable SPECIFIC feature with a [+def]
value or no SPECIFIC feature/value
When the classifier contains no SPECIFIC feature/value, the iSPECIFIC [ ] feature on
the S head remains unvalued. This state of affairs will cause a crash at LF unless there is
a default rule to assign a value to the S head for interpretation. I suggest that a default
rule at LF assigns an indefinite interpretation to an unvalued S head. The rule can be
stated as follows:
(38) An unvalued S head is assigned an indefinite interpretation at LF.
In order to make sure an uninterpretable feature will be matched with an
interpretable one, following Pesetsky & Torrego (2004), I adopt Brody’s (1997) Thesis
of Radical Interpretation, which requires all syntactic elements to be semantically
interpretable. This will force all the uninterpretable features to match with an interpretable
counterpart. Pesetsky & Torrego (2004) notes that Brody’s (1997) Thesis of Radical
Interpretation differs from the deletion of uninterpretable features after checking in that
5
There is at least one Chinese dialect which seems to support the existence of the ‘definite’ and
the ‘indefinite’ type of classifiers. In Wenzhou, [Cl-N] phrases can be interpreted as either
definite or indefinite depending on the tone of the classifier (Cheng & Sybesma 2005). When
the classifier bears its original tone, the interpretation of the [Cl-N] phrase is indefinite; when
the classifier changes into a dipping tone, the [Cl-N] phrase is definite:
(i) dou31 kau35
(ii) dou212 kau35
CL
dog Æ ‘a dog’
CL
dog Æ ‘the dog’
116
The Encoding of Referential Properties in the Chinese Nominal
“it is not uninterpretable features that delete at the interface with the semantic
component—because there can be no uninterpretable features at the semantic interface.
There are only uninterpretable instances of features, and every feature must have at
least one interpretable instance” (Pesetsky & Torrego 2004:8). This implies that when
the SP layer is not projected, the classifier has to be of the type that contains no
feature/value because otherwise the uninterpretable instance of the SPECIFIC feature
cannot be matched with an interpretable instance of the SPECIFIC feature.
The relationship between the S head and the classifier head is illustrated as follows:
(39) a.
Agree
S
… Cl
iSPECIFIC [ ] uSPECIFIC [+def]
b.
S
…
iSPECIFIC [ ]
S
…
Cl
iSPECIFIC [+def]
uSPECIFIC [+def]
No Agree
Cl
∅
S
iSPECIFIC [ ]
…
Cl
∅
I assume that when there is an Agree relation between the S head and the Cl head,
the Cl (both the feature and the category) needs to move to the S head (different from
Pesetsky & Torrego 2004). This assumption is important in accounting for why
[Nume-Cl-N] phrases are always indefinite and the requirement for the definite classifier
to be overt in licensing bare modifiers, which will be made clear in §6 and §7 respectively.
With all these tools in place, we can now proceed to the structures of different noun
phrases.
6. The structures of different noun phrases
In this section, I shall provide the structures of different noun phrases and show
how their interpretations are derived. Since the focus is on the interaction between the
SP layer and the ClP layer, I would only focus on noun phrases where the SP layer is
projected. In other words, only specific noun phrases will be discussed.
6.1 [Cl-N] phrases
[Cl-N] noun phrases can be either definite or indefinite. In the former case, the
classifier comes with a uSPECIFIC [+def] feature, as in (40a); in the latter case, the
classifier comes with no SPECIFIC feature/value, ∅, as in (40b). For ease of exposition,
the SPECIFIC feature will be represented as ‘S’ in the following discussion.
117
Joanna Ut-seong Sio
(40) a.
SP
3
b.
S´
3
ClP
Cli-S
uS[+def]i -iS [+def] 3
Cl´
3
NP
ti
SP
3
S´
3
S
ClP
iS [ ]
3
Cl´
3
Cl
NP
∅
In (40a), the iS [ ] feature on S is the probe and the uS [+def] feature is the goal.
The iS [ ] feature on the S head agrees with the uS [+def] feature on the classifier. The
classifier, with its uS [+def] feature, moves to the S head. The noun phrase is interpreted
as definite. In (40b), there is no Agree. The S head remains unspecified in syntax. The S
head gets assigned an ‘indefinite’ interpretation at LF by the default rule stated in (38).
6.2 [Nume-Cl-N] phrases
[Nume-Cl-N] phrases are always indefinite. In other words, the numeral is
incompatible with a classifier with a [+def] value. As a result, the S head in a [Nume-Cl-N]
phrase can never receive a [+def] value. The ‘incompatibility’ can be made more explicit
derivationally. Consider the structure in (41):
(41) *
118
SP
3
S´
3
S
NumeP
iS [ ]
3
Nume´
3
Nume
ClP
3
Cl´
3
Cl
uS [+def]
The Encoding of Referential Properties in the Chinese Nominal
The iS [ ] feature on the S head probes. The classifier, with its uS [+def] feature,
has to move up to the S head. I assume that the overt numeral will block the movement
of the classifier. In other words, the presence of an overt numeral head is incompatible
with a [+def] classifier, accounting for the indefinite interpretation of [Nume-Cl-N]
phrases. When the classifier contains no feature/value, there is no Agree. The classifier
does not have to move and is thus compatible with the presence of the numeral. This is
illustrated below:
(42)
SP
3
S´
3
S
NumeP
iS [ ]
3
Nume´
3
Nume
ClP
3
Cl´
3
Cl
Ø
The noun phrase is interpreted as specific indefinite.
6.3 [Dem-(Nume)-Cl-N] phrases
Noun phrases containing the demonstrative are always definite, even in the
presence of the numeral. Assuming that numerals are not compatible with classifiers
with a uS [+def] feature. The definiteness of the [Dem-Nume-Cl-N] phrases has to
come from a source other than the classifier. I assume that demonstratives are inherently
definite, so they carry a iS [+def] feature. When a demonstrative is merged in the S head,
the S head receives a [+def] value from the demonstrative. Structurally, it can be
represented below:
119
Joanna Ut-seong Sio
(43)
SP
3
S´
3
Dem-S
NumeP
iS [+def]-iS [+def] 3
Nume´
3
Nume
ClP
3
Cl´
3
Cl
Ø
When the numeral is absent, the demonstrative is compatible with both types of
classifiers, one with a uS[+def] feature and one with no feature/value ∅. When the
classifier has a uS[+def] feature, the classifier, with its uS [+def] feature will move to S
to agree with the iS [ ] feature on the S head. The iS [+def] feature on the demonstrative
will also agree with iS [ ] feature on the S head. This is shown in (44a).
(44) a.
SP
3
S´
3
ClP
Dem-Cli-S
iS [+def] -uS [+def]i -iS [+def] 3
Cl´
3
NP
ti
When the classifier has no feature/value, only the iS [+def] feature on the
demonstrative agrees with iS [ ] feature on the S head. This is illustrated in (44b).
120
The Encoding of Referential Properties in the Chinese Nominal
(44) b.
SP
3
S´
3
Dem-S
iS [+def] -iS [+def]
ClP
3
Cl´
3
NP
Cl
Ø
6.4 Bare nouns
As noted earlier, I assume that the Classifier Phrase is always present in Chinese
bare nouns due to the multi-functions of the classifier, definite and indefinite bare noun
phrases will have the following structures:
(45) a.
a definite bare noun
b. an indefinite bare noun
SP
SP
3
3
S´
S´
3
3
ClP
S
ClP
Cli- S
iS [ ]
3
uS [+def]i -iS [+def] 3
Cl´
Cl´
3
3
NP
Cl
NP
ti
∅
When the classifier is definite, as in (45a), the classifier, with its uS [+def] feature,
has to move to the S head. Again, when the classifier has no feature, as in (45b), no Agree
takes place and the S head remains unvalued and consequently receives an indefinite
interpretation from LF.
7. The licensing of modifiers
The impetus for developing a structural account for the encoding of referential
properties in Chinese comes from the interaction between modifier positions and
121
Joanna Ut-seong Sio
referential properties. In what follows, I would discuss how the licensing of modifiers in
the referential domain can be cast in structural terms in accordance with the theory
proposed here.
Recall that the following observations were previously made:
(46) a.
Only a specific [Nume-Cl-N] phrase would allow a marker modifier
appearing to its immediate left.
b. Only a definite [Cl-N] phrase would allow a bare modifier appearing to
its immediate left.
The pattern above can be partially explained assuming that (i) that only specific
noun phrases project the SP layer and (ii) modifiers reside in the SP layer. For non-specific
noun phrases, no modifiers can appear to the left of the classifier or the numeral because
the superstructure of the noun phrase ends in the Classifier Phrase or the Numeral Phrase.
With respect to the different types of modifiers, a marker modifier can appear to
the left of a [Nume-Cl-N] phrase as shown in §4.1. A marker modifier can also appear
to the left of a [Dem-(Nume)-Cl-N] phrase, as in (47):
(47) sik6jin1
smoke
ge3
go2
MARKER
DEM
saam1
three
go3
CL
hok6saang1
student
Translated into structural terms, it means that both a definite and an indefinite
specific noun phrase (i.e. an S that has a [+def] value and an S that is valueless) allow a
marker modifier to appear in SpecSP. In other words, marker modifiers are oblivious of
the definiteness value of the S head. Bare modifiers can appear to the left of a
[Dem-(Nume)-Cl-N] and a definite [Cl-N] phrase. However, [bare modifier-Nume-Cl-N]
phrases are bad. In structural terms, it means that only an S head with a [+def] value can
license bare modifiers in SpecSP.6
However, merely definiteness is not enough. As noted earlier on, an overt classifier
is also necessary. Definite bare nouns (e.g. Mandarin) cannot license bare modifiers.
Why is an overt classifier necessary? I would like to suggest that it is an overt S head
that is needed.
If the S head is the licenser for bare modifiers, it is plausible that some requirements would be posted on S in order for it to license bare modifiers in its spec. On the
other hand, it is unclear how the classifier, (which is at least one projection down), is
relevant in the licensing of element in SpecSP. The situation can be made sense of if the
6
[marker modifier-Cl-N] phrases are also horrendously bad. It is unclear to me why it is the
case.
122
The Encoding of Referential Properties in the Chinese Nominal
overt classifier requirement is viewed as an overt S requirement. The shift of perspective is possible as in the present analysis; feature movement has to be accompanied by
category movement (Chomsky 1995). Only when the classifier is definite, will the S
head be filled by an overt classifier. Consequentially, only when the classifier is definite,
are bare modifiers in SpecSP licensed. When the demonstrative is present, however,
then the demonstrative makes the S head overt, and licensing is again possible.
8. Conclusions
In this paper, I proposed a formal account for the encoding of referential properties
in the Chinese nominal. For a certain Chinese noun phrase with a surface sequence X,
each of X’s possible referential interpretations is linked unambiguously to a corresponding underlying structure with a certain feature specification. Different underlying structures with different feature make-up might give rise to identical surface sequences as in
a Chinese nominal, not all functional heads with varying feature specification are provided with distinct morphology. In such cases, X is ambiguity and is compatible with
several interpretations.
123
Joanna Ut-seong Sio
References
Arsenijevic, Boban. 2008. A non-linear discourse representation. Manuscript. Amsterdam:
University of Amsterdam.
Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing.
Chicago: University of Chicago Press.
Borer, Hagit. 2005. Structuring Sense, Vol. 1: In Name Only. Oxford & New York:
Oxford University Press.
Brody, Michael. 1997. Perfect chains. Elements of Grammar: Handbook in Generative
Syntax, ed. by Liliane M. V. Haegeman, 139-167. Dordrecht: Kluwer.
Brugè, Laura. 2002. The positions of demonstratives in the extended nominal projection.
Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol.
1, ed. by Guglielmo Cinque, 15-53. Oxford & New York: Oxford University Press.
Campbell, Richard. 1996. Specificity operator in SpecDP. Studia Linguistica 50.2:
161-188.
Chao, Yuen Ren. 1968. A Grammar of Spoken Chinese. Berkeley: University of California Press.
Cheng, Lisa L.-S., and Rint Sybesma. 1999. Bare and not-so-bare nouns and the
structure of NP. Linguistic Inquiry 30.4:509-542.
Cheng, Lisa L.-S., and Rint Sybesma. 2005. Classifiers in four varieties of Chinese. The
Oxford Handbook of Comparative Syntax, ed. by Guglielmo Cinque & Richard S.
Kayne, 259-292. Oxford & New York: Oxford University Press.
Cheung, Samuel H.-N. 1989. Yueyu liangci yongfa de yanjiu [A study of the use of
classifiers in Cantonese]. Papers from the Second International Conference on
Sinology (Linguistics and Philology Section), 753-774. Taipei: Academia Sinica.
Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge: MIT Press.
Chomsky, Noam. 2000. Minimalist inquiries: the framework. Step by Step: Essays on
Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin, David
Michaels & Juan Uriagereka, 89-155. Cambridge: MIT Press.
Chomsky, Noam. 2001. Derivation by phase. Ken Hale: A Life in Language, ed. by
Michael J. Kenstowicz, 1-52. Cambridge: MIT Press.
Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP.
Paths Towards Universal Grammar: Studies in Honor of Richard S. Kayne, ed. by
Guglielmo Cinque et al., 85-110. Washington, DC: Georgetown University Press.
Emonds, Joseph E. 1976. A Transformational Approach to English Syntax: Root,
Structure-preserving, and Local Transformations. New York: Academic Press.
124
The Encoding of Referential Properties in the Chinese Nominal
Emonds, Joseph E. 1978. The verbal complex V’-V in French. Linguistic Inquiry 9.2:
151-175.
Hoekstra, Teun, and Nina Hyams. 1996. The syntax and interpretation of dropped
categories in child language: a unified account. Proceedings of the WCCFL-14,
123-136. Stanford: CSLI.
Huang, C.-T. James. 1982. Logical Relations in Chinese and the Theory of Grammar.
Cambridge: MIT dissertation.
Lee, Thomas H.-T. 1986. Studies on Quantification in Chinese. Los Angeles: University
of California dissertation.
Li, Charles N., and Sandra A. Thompson. 1981. Mandarin Chinese: A Functional
Reference Grammar. Berkeley: University of California Press.
Lyons, Christopher. 1999. Definiteness. Cambridge & New York: Cambridge University
Press.
Pesetsky, David, and Esther Torrego. 2004. The syntax of valuation and the interpretability of features. Manuscript. Cambridge: MIT; Boston: University of Massachusetts. http://web.mit.edu/linguistics/people/faculty/pesetsky/Pesetsky_Torrego_
Agree_paper.pdf
Pollock, Jean-Yves. 1989. Verb movement, universal grammar and the structure of IP.
Linguistic Inquiry 20.3:365-424.
Szabolcsi, Anna. 1994. The noun phrase. Syntax and Semantics, Vol. 27: The Syntactic
Structure of Hungarian, ed. by Ferenc Kiefer & Katalin E. Kiss, 179-274. San
Diego: Academic Press.
Zhang, Niina. 2006. Representing specificity by the internal order of indefinites.
Linguistics 44.1:1-21.
[Received 29 January 2007; revised 17 December 2007; accepted 24 December 2007]
Department of Chinese and Bilingual Studies
Hong Kong Polytechnic University
Hung Hom, Kowloon
Hong Kong
[email protected]
125
Joanna Ut-seong Sio
漢語名詞指稱特徵的編碼
蕭月嫦
香港理工大學
從對修飾語的位置的研究可以發現，漢語中類似專指 (specificity) 和定指
(definiteness) 等指稱性特徵與數詞之上的某個結構密切相關。這個結構被稱
為專指短語，這個短語的投射能產生專指的意義，中心語 S 的特徵決定一個
名詞短語的定指意義。並非所有的名詞短語都可以被看作有定的。對定指意
義的限制可以從理論上加以闡釋，定指意義是由專指短語和量詞短語之間的
一致 (Agree) (Chomsky 2000, Pesetsky & Torrego 2004) 關係決定的。每個指稱
意義都一定與帶有某種特徵的相應名詞短語相關聯。
關鍵詞：定指，專指，量詞，修飾語
126